The problem is that string theories are potentially verifiable by experiment, so they are not metaphysical is the most abstract sense. They meet the basic structure of a scientific theory that is subject to falsification. But the energy required for most obvious experimental proofs is so high that they are practically unverifiable.
So how to choose meta-theories like string theory or its alternatives in the meantime. Some argue that mathematical beauty and the ability to verify further mathematical understandings of the theory is a form of proof. But it's not clear what is gained in such situations other than a pretty theory. Which for some folks is enough.
-- Nathan Newman